2006ApJ...645.1076N

We show that in order to minimize the uncertainties in the N and O abundances of low-mass, low-metallicity (O/H≤1/5 solar) emission-line galaxies, it is necessary to employ separate parameterizations for inferring T_e_(N⁺) and T_e(O⁺) from T_e(O⁺²). In addition, we show that for the above systems, the ionization correction factor (ICF) for obtaining N/O from N⁺/O⁺, where the latter is derived from optical emission-line flux ratios, is ICF=1.08±0.09. These findings are based on state-of-the-art single-star H II region simulations, employing our own modeled stellar spectra as input. Our models offer the advantage of having matching stellar and nebular abundances. In addition, they have O/H as low as 1/50 solar (lower than any past work), as well as log(N/O) and log(C/O) fixed at characteristic values of -1.46 and -0.7, respectively. The above results were used to rederive N and O abundances for a sample of 68 systems with 12+log(O/H)≤8.1, whose dereddened emission-line strengths were collected from the literature. The analysis of the log(N/O) versus 12+log(O/H) diagram of the above systems shows that (1) the largest group of objects forms the well-known N/O plateau with a value for the mean (and its statistical error) of -1.43^+0.0084_–0.0085, (2) the objects are distributed within a range in log(N/O) of -1.54 to -1.27 in Gaussian fashion around the mean with a standard deviation of σ=^+0.071_–0.084, and (3) a χ² analysis suggests that only a small amount of the observed scatter in log(N/O) is intrinsic.